Exporatory Data Analysis of NEFSC Bottom Trawl Survey Data

Exploration of spatial and temporal patterns in abundance, and bodymass of fishes from the Northeast groundfish survey. Build code containing data wrangling and conversions can be accessed here.

# Do some formatting
weights_20 <- weights_20 %>% 
  mutate(
    id = as.character(id),
    season = ifelse(season %in% c("SPRING", "Spring"), "Spring", "Fall"),
    season = factor(season, levels = c("Spring", "Fall"))
  )

# Run Summary Functions
ann_means <- ss_annual_summary(weights_20) %>% 
  mutate(`Research Vessel` = ifelse(est_year > 2008, "HB", "AL"))
seasonals <- ss_seasonal_summary(weights_20) %>% 
  mutate(`Research Vessel` = ifelse(est_year > 2008, "HB", "AL"))

# bind them so you can facet
summs <- bind_rows(ann_means, seasonals) %>% 
  mutate(season = factor(season, levels = c("Spring", "Fall", "Spring + Fall")))


#drop haddock to see if that changes the bump
haddock_ann  <- ss_annual_summary(filter(weights_20, comname != "haddock"))
haddock_seas <- ss_seasonal_summary(filter(weights_20, comname != "haddock"))
no_haddock <- bind_rows(haddock_ann, haddock_seas) %>% 
  mutate(season = factor(season, levels = c("Spring", "Fall", "Spring + Fall")))

Spatial Patterns

For large regions like Georges Bank and the Gulf of Maine, what kind of patterns are we seeing.

# Load the strata
survey_strata <- read_sf(str_c(res_path, "Shapefiles/BottomTrawlStrata/BTS_Strata.shp"))  %>% 
  clean_names() %>% 
  filter(strata >= 01010 ,
         strata <= 01760,
         strata != 1310,
         strata != 1320,
         strata != 1330,
         strata != 1350,
         strata != 1410,
         strata != 1420,
         strata != 1490) 


# Key to which strata = which regions
strata_key <- list(
  "Georges Bank"          = as.character(13:23),
  "Gulf of Maine"         = as.character(24:40),
  "Southern New England"  = str_pad(as.character(1:12), width = 2, pad = "0", side = "left"),
  "Mid-Atlantic Bight"    = as.character(61:76))


# Assign Areas
survey_strata <- survey_strata %>% 
  mutate(
    strata = str_pad(strata, width = 5, pad = "0", side = "left"),
    strata_num = str_sub(strata, 3, 4),
    area = case_when(
      strata_num %in% strata_key$`Georges Bank` ~ "Georges Bank",
      strata_num %in% strata_key$`Gulf of Maine` ~ "Gulf of Maine",
      strata_num %in% strata_key$`Southern New England` ~ "Southern New England",
      strata_num %in% strata_key$`Mid-Atlantic Bight` ~ "Mid-Atlantic Bight",
    TRUE ~ "Outside Major Study Areas"
  )) %>% 
  select(finstr_id, strata, strata_num, area, a2, str2, set, stratuma, str3, geometry)

# Load new england
new_england <- ne_states("united states of america") %>% st_as_sf(crs = 4326) 
canada <- ne_states("canada") %>% st_as_sf(crs = 4326) 


# Make trawl data an sf dataset
trawl_sf <- weights_20 %>% st_as_sf(coords = c("decdeg_beglon", "decdeg_beglat"), crs = 4326)

Trawl Regions

# Plot to check
ggplot() +
  geom_sf(data = new_england) +
  geom_sf(data = canada) +
  geom_sf(data = survey_strata, aes(fill = area)) +
  coord_sf(xlim = c(-77, -65.5), ylim = c(34, 45.75), expand = FALSE) +
  guides(fill = guide_legend(nrow = 2)) +
  theme_bw() +
  theme(legend.position = "bottom", legend.title = element_blank())

Regional Summaries

# Just Area, all seasona
area_summs <- weights_20 %>% 
  group_by(survey_area) %>% 
  summarise(
    season = "Spring + Fall",
    lw_biomass_kg = sum(sum_weight_kg, na.rm = T),
    n_stations = n_distinct(id),
    lw_biomass_per_station = lw_biomass_kg / n_stations,
    mean_ind_bodymass = weighted.mean(ind_weight_kg, weights = numlen_adj),
    mean_ind_length = weighted.mean(length, weights = numlen_adj)
  )

# Area x Season
seas_area <- weights_20 %>% 
  group_by(survey_area, season) %>% 
  summarise(
    lw_biomass_kg = sum(sum_weight_kg, na.rm = T),
    n_stations = n_distinct(id),
    lw_biomass_per_station = lw_biomass_kg / n_stations,
    mean_ind_bodymass = weighted.mean(ind_weight_kg, weights = numlen_adj),
    mean_ind_length = weighted.mean(length, weights = numlen_adj)
  )


# Combine those two
summs_combined <- bind_rows(area_summs, seas_area) %>% 
  mutate(season = factor(season, levels = c("Spring", "Fall", "Spring + Fall")))

summs_combined %>% 
  mutate_if(is.numeric,round, 2) %>% 
  arrange(survey_area,season) %>% 
  knitr::kable()
survey_area season lw_biomass_kg n_stations lw_biomass_per_station mean_ind_bodymass mean_ind_length
GB Spring 4183904 2607 1604.87 0.86 39.55
GB Fall 9044802 2343 3860.35 0.70 37.14
GB Spring + Fall 13228706 4950 2672.47 0.78 38.36
GoM Spring 2779467 3005 924.95 0.63 34.55
GoM Fall 7022811 2952 2379.00 0.68 35.81
GoM Spring + Fall 9802278 5957 1645.51 0.65 35.20
MAB Spring 5703280 1995 2858.79 0.84 44.38
MAB Fall 1176131 1925 610.98 0.65 25.67
MAB Spring + Fall 6879411 3920 1754.95 0.78 38.43
SNE Spring 3256620 2237 1455.80 0.56 36.25
SNE Fall 1736036 2093 829.45 0.49 33.04
SNE Spring + Fall 4992656 4330 1153.04 0.53 34.99
# Year x Area
area_summs_y <- weights_20 %>% 
  group_by(est_year, survey_area) %>% 
  summarise(
    season = "Spring + Fall",
    lw_biomass_kg = sum(sum_weight_kg, na.rm = T),
    n_stations = n_distinct(id),
    lw_biomass_per_station = lw_biomass_kg / n_stations,
    mean_ind_bodymass = weighted.mean(ind_weight_kg, weights = numlen_adj),
    mean_ind_length = weighted.mean(length, weights = numlen_adj)
  )

Total Biomass

area_summs_y %>% 
  ggplot(aes(est_year, lw_biomass_kg)) +
    geom_line() +
    facet_wrap(~survey_area, ncol = 2) +
    scale_y_continuous(labels = scales::comma_format()) +
    labs(x = "", y = "Total Biomass (kg)")

CPUE

area_summs_y %>% 
  ggplot(aes(est_year, lw_biomass_per_station)) +
    geom_line() +
    facet_wrap(~survey_area, ncol = 2) +
    labs(x = "", y = "Adjusted Biomass per Station (kg)")

Comparisons to Older Data

Concerns have been raised that the datasets obtained through the NEFSC are inconsistent in some areas over time. The following plots seek to identify differences in a dataset obtained in 2016 from what we currently are exploring with the 2020 dataset.

Each file was processed for size-spectra analysis using the same processing steps. This includes the same species codes, the same abundance and stratification adjustments, and the same L-W derived biomasses.

weights_16 <- read_csv(here::here("data/ss_prepped_data/survdat_2016_ss.csv"),
                       col_types = cols(),
                       guess_max = 1e5)
weights_19 <- read_csv(here::here("data/ss_prepped_data/survdat_2019_ss.csv"),
                       col_types = cols(),
                       guess_max = 1e5)
weights_20 <- read_csv(here::here("data/ss_prepped_data/survdat_2020_ss.csv"),
                       col_types = cols(),
                       guess_max = 1e5)

# run summaries
summ_16 <- ss_regional_differences(weights_16) %>% mutate(source = "2016")
summ_19 <- ss_regional_differences(weights_19) %>% mutate(source = "2019")
summ_20 <- ss_regional_differences(weights_20) %>% mutate(source = "2020")
reg_summs <- bind_rows(list(summ_16, summ_19, summ_20))
# Total Biomass
p1 <- reg_summs %>% 
  ggplot(aes(est_year, lw_biomass_kg, color = source)) +
  geom_line(show.legend = F) +
  scale_y_continuous(labels = scales::comma_format()) +
  facet_wrap(~survey_area, ncol = 1, scales = "free") +
  labs(x = "", y = "Total Biomass \n (L-W Regressions)")

# Total Biomass - FSCS
p2 <- reg_summs %>% 
  ggplot(aes(est_year, fscs_biomass_kg, color = source)) +
  geom_line() +
  scale_y_continuous(labels = scales::comma_format()) +
  facet_wrap(~survey_area, ncol = 1) +
  labs(x = "", y = "Total Biomass \n (FSCS Haul Weights)")

# effort
p3 <- reg_summs %>% 
  ggplot(aes(est_year, n_stations, color = source)) +
  geom_line(show.legend = F) +
  scale_y_continuous(labels = scales::comma_format()) +
  facet_wrap(~survey_area, ncol = 1) +
  labs(x = "", y = "Effort (haul count)")

# Species 
p4 <- reg_summs %>% 
  ggplot(aes(est_year, n_species, color = source)) +
  geom_line(show.legend = F) +
  scale_y_continuous(labels = scales::comma_format()) +
  facet_wrap(~survey_area, ncol = 1) +
  labs(x = "", y = "Distinct Species")

p1 + p2 + p3 + p4

 

A work by Adam A. Kemberling

Akemberling@gmri.org